An adaptive neural network control method for hydraulic manipulator

Through the adaptive neural network, the dynamic model and state evaluation of hydraulic robot arm are simulated, and the control strategy is optimized, which solves the delay problem of hydraulic robot arm in different environments and tasks, improves real-time operation accuracy and adaptability, and realizes intelligence control.

CN118848994BActive Publication Date: 2024-11-22NORTHEASTERN UNIV CHINA +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411345571.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-26
Publication Date
2024-11-22
Estimated Expiration
2044-09-26

AI Technical Summary

Technical Problem

When hydraulic robotic arms face different working environments and tasks, the neural network adjustment process causes delays, affecting real-time operation accuracy.

Method used

By acquiring and preprocessing the operation data and task data of the hydraulic robot arm, using an adaptive neural network to simulate the dynamic model of the robot arm, combining historical data to build an operating state evaluation model, optimize neural network parameters, adjust control strategies in real time, and generate adaptive Control command.

Benefits of technology

It improves the control accuracy and adaptability of the robotic arm, reduces task interruption or failure, improves task execution efficiency and success rate, and realizes intelligent control.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118848994B_ABST
    Figure CN118848994B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of intelligent robot mechanical arm control and regulation technology. The present invention provides a hydraulic mechanical arm adaptive neural network control method, which obtains the hydraulic mechanical arm operation and task data, uses an adaptive neural network to simulate the mechanical arm dynamics model to predict the dynamic response, and combines historical data to build an operation status evaluation model. Through real-time error analysis and data splitting, the model is optimized to improve the prediction accuracy, real-time data is matched according to historical usage scenarios, task parameters are corrected, and adaptive control commands are generated. The present invention significantly improves the control accuracy, environmental adaptability and task execution efficiency of the hydraulic mechanical arm.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent robot mechanical arm control and regulation, and in particular to an adaptive neural network control method for a hydraulic mechanical arm. Background Art

[0002] The hydraulic manipulator is mainly composed of hydraulic cylinder, hydraulic pump, hydraulic valve, control system and actuator (i.e. the manipulator itself). The working principle of the hydraulic manipulator is to provide power through the hydraulic pump to drive the piston in the hydraulic cylinder to drive the manipulator to perform various actions. The control system of the hydraulic manipulator is responsible for accurately controlling the movement of the hydraulic cylinder according to the preset program or real-time sensor data to achieve precise operation of the manipulator. Adaptive neural network is a neural network that can self-adjust to adapt to different environments and tasks. The characteristic of this network is that it can automatically adjust its structure, parameters or learning strategy according to the characteristics of the input data and the task requirements to improve performance and efficiency.

[0003] The movement of the robot is a highly nonlinear process, and traditional control methods often have difficulty accurately describing its dynamic characteristics. Adaptive neural networks can use their nonlinear approximation capabilities to simulate the dynamic model of the robot. Through a large amount of training data, the neural network can learn the movement law of the robot, thereby accurately predicting the dynamic response of the robot under different states. Since the state of the robot is difficult to measure directly, the adaptive neural network can be combined with a high-gain observer to estimate the state of the system. This method achieves accurate estimation of the state of the robot by comparing the observation error with the output of the neural network, and then performs precise control.

[0004] When adaptive neural networks are applied to intelligent robot arms, there are the following technical pain points. The robot's hydraulic arm needs to face different working environments and tasks. The movement of the hydraulic arm is a highly nonlinear process, and its state is difficult to measure directly. Although the adaptive neural network can be combined with a high-gain observer to estimate the state of the system, when facing different working environments and tasks to handle control problems with high uncertainty, the adjustment process of the neural network will cause a certain delay, which will affect the real-time operation accuracy of the hydraulic arm. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an adaptive neural network control method for a hydraulic mechanical arm to solve the problem that the hydraulic mechanical arm of a robot needs to face different working environments and tasks. When facing different working environments and tasks and handling control problems with high uncertainty, the adjustment process of the neural network will cause a certain delay, and the delay will affect the real-time operation accuracy of the hydraulic mechanical arm.

[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0007] The present invention provides a hydraulic mechanical arm adaptive neural network control method, comprising:

[0008] Step S101, obtaining hydraulic manipulator arm operation data and hydraulic manipulator arm task data, the hydraulic manipulator arm operation data includes hydraulic manipulator arm state data, hydraulic manipulator arm position, hydraulic manipulator arm speed and hydraulic manipulator arm acceleration, and the hydraulic manipulator arm task data includes target position, target posture, grasped object position, hydraulic manipulator arm trajectory planning parameters, hydraulic manipulator arm target torque, task constraint time and sensor data requirements;

[0009] Step S102, preprocessing the hydraulic mechanical arm operation data and the hydraulic mechanical arm task data, using the preprocessed hydraulic mechanical arm operation data and the hydraulic mechanical arm task data to simulate the mechanical arm dynamics model, and predicting the dynamic response data of the mechanical arm in different states, to obtain the action prediction data of the mechanical arm in different states;

[0010] Step S103, retrieve the historical operation data and historical task data of the hydraulic mechanical arm, build a hydraulic mechanical arm operation status evaluation model according to the historical operation data and historical task data of the hydraulic mechanical arm, obtain the real-time hydraulic mechanical arm operation data and real-time hydraulic mechanical arm task data, substitute the real-time hydraulic mechanical arm task data into the hydraulic mechanical arm operation status evaluation model, obtain the hydraulic mechanical arm operation status evaluation result, compare the hydraulic mechanical arm operation status evaluation result and the action prediction data of the mechanical arm in different states with the real-time hydraulic mechanical arm operation data, and obtain the real-time hydraulic mechanical arm operation error data;

[0011] Step S104, splitting the real-time hydraulic mechanical arm operation error data according to the task type to obtain error data corresponding to each task, combining the error data corresponding to each task with the hydraulic mechanical arm historical operation data and the hydraulic mechanical arm historical task data in turn to obtain a data group to be optimized, and using the data combination in the data group to be optimized in turn to optimize the adaptive neural network simulation mechanical arm dynamics model and the hydraulic mechanical arm operation status evaluation model to obtain an optimized adaptive neural network simulation mechanical arm dynamics model combination and an optimized hydraulic mechanical arm operation status evaluation model combination;

[0012] Step S105, identify the historical operation data and historical task data of the hydraulic mechanical arm, obtain the historical hydraulic mechanical arm usage scenario type, process the real-time hydraulic mechanical arm operation data according to the historical hydraulic mechanical arm usage scenario type, obtain the corrected hydraulic mechanical arm task data, and generate the hydraulic mechanical arm adaptive control command according to the corrected hydraulic mechanical arm task data.

[0013] Furthermore, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S101, comprises:

[0014] A hydraulic mechanical arm data category is established based on historical hydraulic mechanical arm operation data and historical hydraulic mechanical arm task data. The hydraulic mechanical arm data category includes a hydraulic mechanical arm task category and hydraulic mechanical arm operation data corresponding to each hydraulic mechanical arm task category.

[0015] Furthermore, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S102 includes:

[0016] The adaptive neural network model is trained using historical hydraulic manipulator operation data and task data, and the pre-processed real-time hydraulic manipulator operation data and task data are input into the trained adaptive neural network simulation manipulator dynamics model. The adaptive neural network simulation manipulator dynamics model predicts the dynamic response data of the manipulator under different states based on the input data. The dynamic response data under different states include hydraulic manipulator position prediction data, hydraulic manipulator speed prediction data, and hydraulic manipulator acceleration prediction data.

[0017] The motion prediction data of the robot arm in different states are obtained from the output layer of the adaptive neural network simulation robot arm dynamics model. The motion prediction data of the robot arm in different states are used for the operation state evaluation of the hydraulic robot arm and the correction of the task data.

[0018] Furthermore, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S103 includes:

[0019] The hydraulic manipulator arm historical operation data includes the hydraulic manipulator arm historical position, hydraulic manipulator arm historical speed, hydraulic manipulator arm historical force, hydraulic manipulator arm historical acceleration and hydraulic manipulator arm historical environmental parameters, and the hydraulic manipulator arm historical task data includes detailed information of each task of the hydraulic manipulator arm;

[0020] The historical operation data and historical task data of the hydraulic mechanical arm are standardized according to the timestamp to obtain the historical operation data and historical task data of the hydraulic mechanical arm that are consistent in time;

[0021] The average operation time characteristics, maximum load characteristics, and average failure interval characteristics are obtained from the historical operation data of the hydraulic manipulator;

[0022] The historical operation data and historical task data of the hydraulic manipulator are used as training sets, and the recurrent neural network model is trained using the training set data to obtain the hydraulic manipulator operation status evaluation model.

[0023] Furthermore, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S104 comprises:

[0024] By analyzing the identifiers and parameters in the task data, the task type corresponding to the real-time hydraulic manipulator operation data is obtained;

[0025] The real-time hydraulic robot arm operation error data is classified according to the task type. For each task type, the task error data is extracted from the classified error data set, and the error data set corresponding to each task is output. The error data set corresponding to each task also includes a combination of the hydraulic robot arm historical operation data and task data.

[0026] Furthermore, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S104 comprises:

[0027] For each task type, the real-time error data corresponding to each task type is combined with the historical operation data and historical task data of the hydraulic manipulator;

[0028] Arrange the combined data according to a fixed data format and structure to form a data group to be optimized;

[0029] Each data group to be optimized contains the hydraulic manipulator operation data, task data and corresponding error data under a specific task type;

[0030] The adaptive neural network simulation mechanical arm dynamics model is trained using the data combination in the data group to be optimized to obtain the data combination training results in the data group to be optimized, and the parameters of the neural network are adjusted according to the data combination training results in the data group to be optimized to obtain the optimized adaptive neural network simulation mechanical arm dynamics model and the optimized hydraulic mechanical arm operation status evaluation model corresponding to each task;

[0031] The optimized adaptive neural network simulation robot arm dynamics model and the hydraulic robot arm operation status evaluation model are combined to form an optimized model combination.

[0032] Furthermore, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S105 comprises:

[0033] The historical operation data and historical task data of the hydraulic mechanical arm are identified to obtain the historical hydraulic mechanical arm usage scenario types, and a matching relationship is established between the hydraulic mechanical arm operation data corresponding to the historical hydraulic mechanical arm usage scenario types, the optimized adaptive neural network simulation mechanical arm dynamics model, and the optimized hydraulic mechanical arm operation status evaluation model to obtain the hydraulic mechanical arm usage scenario data group, and the real-time hydraulic mechanical arm operation data and the real-time hydraulic mechanical arm task data are matched with the hydraulic mechanical arm usage scenario data group to obtain the real-time hydraulic mechanical arm scene data group, and the real-time hydraulic mechanical arm scene data group is matched with the corresponding adaptive neural network simulation mechanical arm dynamics. The model and the hydraulic mechanical arm operation status evaluation model are used for data analysis to obtain the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario, and the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario are compared with the real-time hydraulic mechanical arm operation data. If the comparison result error is greater than the preset error range, the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario are used to correct the hydraulic mechanical arm task data to obtain the corrected hydraulic mechanical arm task data, and the hydraulic mechanical arm adaptive control command is generated according to the corrected hydraulic mechanical arm task data.

[0034] Furthermore, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S105 comprises:

[0035] Compare the real-time hydraulic mechanical arm action prediction data of each scene, the real-time hydraulic mechanical arm operation state prediction data of each scene and the real-time hydraulic mechanical arm operation data, calculate the error value between the real-time hydraulic mechanical arm action prediction data of each scene and the real-time hydraulic mechanical arm operation data to obtain a first error value, calculate the error value between the real-time hydraulic mechanical arm operation state prediction data of each scene and the real-time hydraulic mechanical arm operation data to obtain a second error value;

[0036] Determine whether the first error value and the second error value are greater than a preset error range. If the error value is greater than the preset range, use the real-time hydraulic mechanical arm action prediction data of each scene and the real-time hydraulic mechanical arm operation status prediction data of each scene as reference data, correct the target position parameters, target posture parameters and grasped object position parameters in the hydraulic mechanical arm task data to obtain the corrected hydraulic mechanical arm task data, recalculate the control parameters of the hydraulic mechanical arm according to the corrected hydraulic mechanical arm task data to obtain the corrected control parameters, and use the corrected control parameters to generate an adaptive control command for the hydraulic mechanical arm.

[0037] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0038] The present invention simulates the mechanical arm dynamics model through an adaptive neural network, and can more accurately predict the dynamic response of the mechanical arm in different states, thereby improving the control accuracy of the mechanical arm and enabling the mechanical arm to perform predetermined tasks more accurately.

[0039] The present invention utilizes historical operation data and task data to train and optimize the adaptive neural network, so that it can automatically adjust the control strategy according to different environments and tasks, thereby enhancing the adaptability of the robot arm to different working environments and tasks.

[0040] Through the analysis and processing of real-time operation error data, the present invention can promptly detect and correct the deviation of the robot arm during operation, thereby reducing task interruption or failure caused by inaccurate operation and improving the efficiency and success rate of task execution.

[0041] The present invention combines historical data and real-time data, and realizes intelligent control of the hydraulic mechanical arm through adaptive neural network and state evaluation model. This control method can dynamically adjust the control parameters according to the real-time state of the mechanical arm and task requirements, thereby improving the flexibility and intelligence level of the control.

[0042] In summary, the hydraulic mechanical arm adaptive neural network control method of the present invention shows significant beneficial effects in improving control accuracy, enhancing environmental adaptability, improving task execution efficiency, realizing intelligent control, reducing maintenance costs and improving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0044] Figure 1 Schematic diagram of the adaptive neural network control method for a hydraulic mechanical arm provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0046] In order to better understand the purpose of the present invention, the present invention is described in further detail below.

[0047] The present invention provides a hydraulic mechanical arm adaptive neural network control method, comprising:

[0048] Step S101, obtaining hydraulic manipulator arm operation data and hydraulic manipulator arm task data, the hydraulic manipulator arm operation data includes hydraulic manipulator arm state data, hydraulic manipulator arm position, hydraulic manipulator arm speed and hydraulic manipulator arm acceleration, and the hydraulic manipulator arm task data includes target position, target posture, grasped object position, hydraulic manipulator arm trajectory planning parameters, hydraulic manipulator arm target torque, task constraint time and sensor data requirements;

[0049] The hydraulic manipulator state data includes the current working state of the manipulator, such as whether it is in working state, fault state, standby state, etc. The hydraulic manipulator position refers to the current position coordinates of the manipulator in three-dimensional space. The hydraulic manipulator speed indicates the moving speed of each joint or end effector of the manipulator. The hydraulic manipulator acceleration describes the acceleration change during the movement of the manipulator, which reflects an important indicator of the dynamic performance of the manipulator.

[0050] The hydraulic manipulator task data includes the target position, which specifies the final position coordinates that the manipulator needs to reach. The target posture describes the specific posture or angle that the manipulator should take when reaching the target position. The position of the grasped object. If the task involves a grasping operation, this will be the initial position of the grasped object. The hydraulic manipulator trajectory planning parameters include the motion path planning parameters of the manipulator from the starting position to the target position, such as path points, speed curves, etc. The hydraulic manipulator target torque is the target torque value that each joint of the manipulator needs to achieve when performing the task. The task constraint time is the time limit required to complete the task, which helps to optimize the motion trajectory and speed of the manipulator. The sensor data requirements are the sensor data required to perform the task, such as force sensors, position sensors, etc., to monitor and feedback the operating status of the manipulator.

[0051] By obtaining the hydraulic robot arm operation data and hydraulic robot arm task data, a solid data foundation can be provided for subsequent steps (such as neural network simulation, state evaluation, error calculation, etc.), thereby improving the hydraulic robot arm's ability to complete tasks accurately and efficiently.

[0052] Step S102, preprocessing the hydraulic mechanical arm operation data and the hydraulic mechanical arm task data, using the preprocessed hydraulic mechanical arm operation data and the hydraulic mechanical arm task data to simulate the mechanical arm dynamics model, and predicting the dynamic response data of the mechanical arm in different states, to obtain the action prediction data of the mechanical arm in different states;

[0053] In step S102, the main task is to pre-process the operation data and task data of the hydraulic manipulator, and use the pre-processed data to simulate the dynamic model of the manipulator through an adaptive neural network, so as to predict the dynamic response data of the manipulator in different states. The following is a detailed technical solution for this step:

[0054] The preprocessing of the operation data and task data of the hydraulic manipulator includes the following aspects:

[0055] Data cleaning: Remove noise, outliers or duplicate data points to ensure the accuracy and consistency of input data.

[0056] Data conversion: Convert the original data into a format suitable for neural network processing, such as normalization or standardization, to avoid the impact of different dimensions on neural network training.

[0057] Feature selection: Extract features useful for neural network prediction from raw data, reduce redundant information, and improve prediction efficiency.

[0058] The preprocessed data is fed into an adaptive neural network to simulate the dynamic model of the hydraulic manipulator. The adaptive neural network has the ability to self-adjust and can automatically adjust its structure and parameters according to the characteristics of the input data and the task requirements. During the simulation, the neural network learns the dynamic behavior of the manipulator and is able to predict the dynamic response of the manipulator in different states.

[0059] The adaptive neural network predicts the dynamic response data of the robot arm in different states based on the input operation data and task data. The dynamic response data includes: Position prediction data: predicting the position coordinates of the robot arm at a certain moment in the future.

[0060] Speed ​​prediction data: predict the movement speed of the robot arm in different states. Acceleration prediction data: predict the acceleration change of the robot arm, reflecting the change trend of its dynamic performance.

[0061] The output layer of the adaptive neural network obtains the action prediction data of the robot arm in different states. These action prediction data not only include the position, velocity and acceleration information of the robot arm at a certain moment in the future, but also include the specific action sequence or control instructions that need to be taken to achieve specific task goals.

[0062] Step S102 effectively predicts the dynamic response and action of the hydraulic mechanical arm under different states by combining data preprocessing and adaptive neural network simulation. This process provides an important reference for subsequent real-time control and helps to improve the accuracy and stability of the hydraulic mechanical arm.

[0063] Step S103, retrieve the historical operation data and historical task data of the hydraulic mechanical arm, build a hydraulic mechanical arm operation status evaluation model according to the historical operation data and historical task data of the hydraulic mechanical arm, obtain the real-time hydraulic mechanical arm operation data and real-time hydraulic mechanical arm task data, substitute the real-time hydraulic mechanical arm task data into the hydraulic mechanical arm operation status evaluation model, obtain the hydraulic mechanical arm operation status evaluation result, compare the hydraulic mechanical arm operation status evaluation result and the action prediction data of the mechanical arm in different states with the real-time hydraulic mechanical arm operation data, and obtain the real-time hydraulic mechanical arm operation error data;

[0064] In step S103, the main tasks are to evaluate the operating status of the hydraulic mechanical arm and calculate the real-time operating error. The following is a detailed technical solution for this step:

[0065] Retrieve the historical operation data (such as historical position, speed, force, acceleration, and environmental parameters) and historical task data (detailed information of each task) of the hydraulic manipulator from the storage system. Use the historical operation data and historical task data of the hydraulic manipulator to build an operation status evaluation model of the hydraulic manipulator through machine learning or deep learning methods (such as recurrent neural network RNN). The operation status evaluation model of the hydraulic manipulator can evaluate the operation status of the hydraulic manipulator under a given task, including its performance, stability, and potential failure risks.

[0066] Acquire the hydraulic manipulator's operating data (such as current position, speed, acceleration, etc.) and task data (such as the target position, posture, trajectory planning parameters, etc. of the current task) in real time. These data reflect the hydraulic manipulator's current working status and the task to be performed.

[0067] The task data acquired in real time is substituted into the established hydraulic manipulator operation status evaluation model. The model evaluates the possible operation status of the hydraulic manipulator when performing the task based on the input task data.

[0068] The evaluation results are compared with the real-time operation data, which is the actual performance data of the hydraulic manipulator during the execution of the task. By comparing the evaluation results with the real-time data, the degree of deviation of the hydraulic manipulator during the execution of the task can be understood, that is, the accuracy of the operation status evaluation.

[0069] The evaluation results of the hydraulic manipulator's operating status and the motion prediction data of the manipulator in different states are compared in detail with the real-time hydraulic manipulator's operating data. The purpose of this step is to calculate the real-time hydraulic manipulator's operating error data. The operating error reflects the difference between the hydraulic manipulator's actual execution process and the predicted state, and is an important indicator for measuring control accuracy.

[0070] Based on the comparison results, a real-time hydraulic robot arm operation error data set is generated. These data sets will be used for further optimization and adjustment of the adaptive neural network model and control strategy in subsequent steps.

[0071] In summary, step S103 achieves accurate evaluation of the operating state of the hydraulic manipulator and effective monitoring of the real-time operating error by a series of operations such as retrieving historical data, building an evaluation model, substituting real-time data, comparing the evaluation results with the real-time data, and calculating the operating error. This step provides a reference for subsequent optimization and control command generation.

[0072] Step S104, splitting the real-time hydraulic mechanical arm operation error data according to the task type to obtain error data corresponding to each task, combining the error data corresponding to each task with the hydraulic mechanical arm historical operation data and the hydraulic mechanical arm historical task data in turn to obtain a data group to be optimized, and using the data combination in the data group to be optimized in turn to optimize the adaptive neural network simulation mechanical arm dynamics model and the hydraulic mechanical arm operation status evaluation model to obtain an optimized adaptive neural network simulation mechanical arm dynamics model combination and an optimized hydraulic mechanical arm operation status evaluation model combination;

[0073] In step S104, the real-time hydraulic manipulator operation error data is processed, and the adaptive neural network simulation manipulator dynamics model and the hydraulic manipulator operation state evaluation model are optimized. The following is a detailed technical solution for this step:

[0074] The real-time hydraulic manipulator operation error data is split according to the task type. Since the hydraulic manipulator performs many different types of tasks, each task has different performance requirements and control strategies for the manipulator. Therefore, classifying the error data by task type helps to more accurately analyze the source of the error and improve the control strategy.

[0075] For each task type, the corresponding real-time error data is combined with the historical operation data and historical task data of the hydraulic robot arm. The purpose of this is to make full use of the experience and rules in the historical data, combined with real-time error feedback, to optimize the model more targetedly.

[0076] The combined data is sorted according to a fixed data format and structure to form a data group to be optimized. Each data group to be optimized contains the hydraulic manipulator operation data, task data and corresponding error data under a specific task type. Such a data structure facilitates the subsequent training and optimization of the model.

[0077] The adaptive neural network simulation robot arm dynamics model and the hydraulic robot arm operation status evaluation model are optimized using the data combination in each data group to be optimized. The goal of optimization is to reduce the prediction error of the model and improve the accuracy and generalization ability of the model. This usually involves adjusting the parameters, structure or learning strategy of the model.

[0078] After multiple iterations of optimization, the optimized adaptive neural network simulation robot arm dynamics model combination and the optimized hydraulic robot arm operation status evaluation model combination were obtained. The optimized model combination can better adapt to the hydraulic robot arm control requirements under different task types and improve the operation accuracy and stability of the robot arm.

[0079] Step S104 achieves continuous improvement and enhancement of the adaptive neural network simulation mechanical arm dynamics model and the hydraulic mechanical arm operation status evaluation model through a series of operations such as splitting the real-time error data, combining it with historical data, forming a data group to be optimized, and optimizing the model. This process enables the adaptive neural network simulation mechanical arm dynamics model to more accurately reflect the actual operation status and control requirements of the hydraulic mechanical arm, thereby improving the operating performance and reliability of the mechanical arm.

[0080] Step S105, identify the historical operation data and historical task data of the hydraulic mechanical arm, obtain the historical hydraulic mechanical arm usage scenario type, process the real-time hydraulic mechanical arm operation data according to the historical hydraulic mechanical arm usage scenario type, obtain the corrected hydraulic mechanical arm task data, and generate the hydraulic mechanical arm adaptive control command according to the corrected hydraulic mechanical arm task data.

[0081] In step S105, the main task is to identify the usage scenario types of the hydraulic manipulator based on historical data, and process the real-time operation data based on these scenario types to generate more accurate adaptive control commands. The following is a detailed technical solution for this step:

[0082] First, we conduct an in-depth analysis of the historical operation data and historical task data of the hydraulic manipulator to identify different usage scenario types. These scenario types may be divided based on a variety of factors such as the working environment of the manipulator, the nature of the task, and the object of operation. For example, the scenario types may include precision assembly, heavy object handling, and high temperature environment operations.

[0083] According to the identified historical usage scenario type, the real-time hydraulic manipulator operation data is processed accordingly. This processing may include data filtering, outlier removal, data interpolation and other technologies, with the aim of improving the accuracy and reliability of the data so as to more accurately reflect the current operating status of the hydraulic manipulator.

[0084] Based on the processed real-time operation data and historical usage scenario information, the current hydraulic manipulator task data is corrected. This correction may involve multiple aspects such as target position, posture, trajectory planning parameters, target torque, etc. By adjusting these parameters, the task data can be made more consistent with the actual operation of the current manipulator and the scenario requirements.

[0085] Finally, according to the corrected hydraulic manipulator task data, the optimized adaptive neural network is used to simulate the manipulator dynamics model and operation status evaluation model to generate the adaptive control command of the hydraulic manipulator. The control command will adjust the manipulator's motion trajectory, speed, acceleration and other parameters in real time so that the manipulator can accurately complete the predetermined task.

[0086] In the process of generating control commands, the real-time operating status of the hydraulic manipulator needs to be continuously monitored and compared with the predicted operating status. If a large deviation is found between the actual operating status and the predicted status, the task data may need to be corrected again and the control command regenerated. This process forms a closed-loop feedback control system, which helps to further improve the operating accuracy and stability of the hydraulic manipulator.

[0087] Step S105 realizes precise control of the hydraulic manipulator in different scenarios by identifying the historical usage scenario type, processing real-time operation data, correcting task data, and generating adaptive control commands. This step makes full use of the advantages of the fusion of historical data and real-time data, as well as the powerful prediction and learning capabilities of the adaptive neural network model, providing strong support for the intelligent control of the hydraulic manipulator.

[0088] Specifically, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S101 includes:

[0089] A hydraulic mechanical arm data category is established based on historical hydraulic mechanical arm operation data and historical hydraulic mechanical arm task data. The hydraulic mechanical arm data category includes a hydraulic mechanical arm task category and hydraulic mechanical arm operation data corresponding to each hydraulic mechanical arm task category.

[0090] Specifically, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S102 includes:

[0091] The adaptive neural network model is trained using historical hydraulic manipulator operation data and task data, and the pre-processed real-time hydraulic manipulator operation data and task data are input into the trained adaptive neural network simulation manipulator dynamics model. The adaptive neural network simulation manipulator dynamics model predicts the dynamic response data of the manipulator under different states based on the input data. The dynamic response data under different states include hydraulic manipulator position prediction data, hydraulic manipulator speed prediction data, and hydraulic manipulator acceleration prediction data.

[0092] The motion prediction data of the robot arm in different states are obtained from the output layer of the adaptive neural network simulation robot arm dynamics model. The motion prediction data of the robot arm in different states are used for the operation state evaluation of the hydraulic robot arm and the correction of the task data.

[0093] Specifically, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S103 includes:

[0094] The hydraulic manipulator arm historical operation data includes the hydraulic manipulator arm historical position, hydraulic manipulator arm historical speed, hydraulic manipulator arm historical force, hydraulic manipulator arm historical acceleration and hydraulic manipulator arm historical environmental parameters, and the hydraulic manipulator arm historical task data includes detailed information of each task of the hydraulic manipulator arm;

[0095] The historical operation data and historical task data of the hydraulic mechanical arm are standardized according to the timestamp to obtain the historical operation data and historical task data of the hydraulic mechanical arm that are consistent in time;

[0096] The average operation time characteristics, maximum load characteristics, and average failure interval characteristics are obtained from the historical operation data of the hydraulic manipulator;

[0097] The historical operation data and historical task data of the hydraulic manipulator are used as training sets, and the recurrent neural network model is trained using the training set data to obtain the hydraulic manipulator operation status evaluation model.

[0098] Specifically, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S104 includes:

[0099] By analyzing the identifiers and parameters in the task data, the task type corresponding to the real-time hydraulic manipulator operation data is obtained;

[0100] The real-time hydraulic robot arm operation error data is classified according to the task type. For each task type, the task error data is extracted from the classified error data set, and the error data set corresponding to each task is output. The error data set corresponding to each task also includes a combination of the hydraulic robot arm historical operation data and task data.

[0101] Specifically, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S104 includes:

[0102] For each task type, the real-time error data corresponding to each task type is combined with the historical operation data and historical task data of the hydraulic manipulator;

[0103] Arrange the combined data according to a fixed data format and structure to form a data group to be optimized;

[0104] Each data group to be optimized contains the hydraulic manipulator operation data, task data and corresponding error data under a specific task type;

[0105] The adaptive neural network simulation mechanical arm dynamics model is trained using the data combination in the data group to be optimized to obtain the data combination training results in the data group to be optimized, and the parameters of the neural network are adjusted according to the data combination training results in the data group to be optimized to obtain the optimized adaptive neural network simulation mechanical arm dynamics model and the optimized hydraulic mechanical arm operation status evaluation model corresponding to each task;

[0106] The optimized adaptive neural network simulation robot arm dynamics model and the hydraulic robot arm operation status evaluation model are combined to form an optimized model combination.

[0107] Specifically, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S105 includes:

[0108] The historical operation data and historical task data of the hydraulic mechanical arm are identified to obtain the historical hydraulic mechanical arm usage scenario types, and a matching relationship is established between the hydraulic mechanical arm operation data corresponding to the historical hydraulic mechanical arm usage scenario types, the optimized adaptive neural network simulation mechanical arm dynamics model, and the optimized hydraulic mechanical arm operation status evaluation model to obtain the hydraulic mechanical arm usage scenario data group, and the real-time hydraulic mechanical arm operation data and the real-time hydraulic mechanical arm task data are matched with the hydraulic mechanical arm usage scenario data group to obtain the real-time hydraulic mechanical arm scene data group, and the real-time hydraulic mechanical arm scene data group is matched with the corresponding adaptive neural network simulation mechanical arm dynamics. The model and the hydraulic mechanical arm operation status evaluation model are used for data analysis to obtain the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario, and the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario are compared with the real-time hydraulic mechanical arm operation data. If the comparison result error is greater than the preset error range, the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario are used to correct the hydraulic mechanical arm task data to obtain the corrected hydraulic mechanical arm task data, and the hydraulic mechanical arm adaptive control command is generated according to the corrected hydraulic mechanical arm task data.

[0109] Specifically, the operation details of the hydraulic mechanical arm adaptive neural network control method provided by the present invention in step S105 are as follows:

[0110] First, the historical operation data and historical task data of the hydraulic manipulator are deeply analyzed to identify different usage scenario types, which reflect the operation characteristics and requirements of the manipulator in different working environments.

[0111] The identified historical hydraulic manipulator arm usage scenario types are matched with the corresponding hydraulic manipulator arm operation data, the optimized adaptive neural network simulation manipulator arm dynamics model, and the optimized hydraulic manipulator arm operation status evaluation model. This step aims to build a comprehensive hydraulic manipulator arm usage scenario database so that the corresponding scenarios and models can be quickly matched based on real-time data in the future.

[0112] The real-time operation data and task data of the hydraulic manipulator are obtained, and the real-time operation data and task data of the hydraulic manipulator are matched with the established usage scenario data group to find the usage scenario that best matches the current real-time data characteristics. The purpose of this step is to enable subsequent model analysis and control command generation to be optimized for specific scenarios.

[0113] For each matched real-time hydraulic manipulator scene data set, the corresponding adaptive neural network simulation manipulator dynamics model and hydraulic manipulator operation status evaluation model are used for data analysis. Through analysis, the action prediction data and operation status prediction data of the manipulator in the current scene are predicted, which provides an important basis for subsequent error comparison and task data correction.

[0114] The predicted data of the real-time hydraulic manipulator action and operation status are compared with the real-time hydraulic manipulator operation data. If the error of the comparison result is greater than the preset error range, it means that there may be problems with the current task data or control strategy and correction is required. At this time, the real-time prediction data is used as a reference to adjust the task data of the hydraulic manipulator to reduce the error and improve the operation accuracy.

[0115] After the task data is corrected, the adaptive control command is recalculated and generated according to the corrected hydraulic manipulator task data. The control command will be directly applied to the control system of the hydraulic manipulator to guide the manipulator to operate according to the corrected task requirements.

[0116] Step S105 achieves precise control and dynamic optimization of the hydraulic manipulator in different usage scenarios through a series of operations such as identifying historical usage scenarios, real-time data matching, data analysis and prediction, error comparison and task data correction, and generating adaptive control commands. This process fully utilizes the powerful capabilities of the adaptive neural network and state assessment model to ensure that the manipulator can maintain efficient and accurate operating performance when facing complex and changing working environments and task requirements.

[0117] Specifically, the hydraulic mechanical arm adaptive neural network control method provided by the present invention, step S105 includes:

[0118] Compare the real-time hydraulic mechanical arm action prediction data of each scene, the real-time hydraulic mechanical arm operation state prediction data of each scene and the real-time hydraulic mechanical arm operation data, calculate the error value between the real-time hydraulic mechanical arm action prediction data of each scene and the real-time hydraulic mechanical arm operation data to obtain a first error value, calculate the error value between the real-time hydraulic mechanical arm operation state prediction data of each scene and the real-time hydraulic mechanical arm operation data to obtain a second error value;

[0119] Determine whether the first error value and the second error value are greater than a preset error range. If the error value is greater than the preset range, use the real-time hydraulic mechanical arm action prediction data of each scene and the real-time hydraulic mechanical arm operation status prediction data of each scene as reference data, correct the target position parameters, target posture parameters and grasped object position parameters in the hydraulic mechanical arm task data to obtain the corrected hydraulic mechanical arm task data, recalculate the control parameters of the hydraulic mechanical arm according to the corrected hydraulic mechanical arm task data to obtain the corrected control parameters, and use the corrected control parameters to generate an adaptive control command for the hydraulic mechanical arm.

[0120] Specifically, the present invention makes a fine adjustment and optimization on the control of the hydraulic mechanical arm in step S105. The following is a detailed technical solution of this step:

[0121] First, the action prediction data and operation status prediction data of the real-time hydraulic mechanical arm in each scenario are compared with the actual real-time hydraulic mechanical arm operation data. By calculating the difference between the prediction data and the actual operation data, the first error value and the second error value are obtained:

[0122] First error value: Calculate the error value between the real-time hydraulic mechanical arm action prediction data of each scene and the real-time hydraulic mechanical arm operation data. The first error value reflects the deviation between the predicted action and the actual action.

[0123] Second error value: Calculate the error value between the predicted data of the real-time hydraulic mechanical arm's operating state in each scenario and the real-time hydraulic mechanical arm's operating data. The second error value measures the gap between the predicted state and the actual state.

[0124] Next, the two error values ​​are judged to see whether the first error value and the second error value exceed a preset error range, which is set according to actual application requirements and the accuracy requirements of the robot arm.

[0125] If both error values ​​are within the preset range, it means that the current prediction and control effects are good and no further adjustment is required.

[0126] If any error value exceeds the preset range, it indicates that the current prediction may be biased, or the actual operating conditions are different from the predicted conditions. At this time, the system will take the following measures to correct it:

[0127] Reference data usage: The action prediction data and operation status prediction data of the real-time hydraulic robot arm in each scene are used as important reference information.

[0128] Task data correction: Based on the above prediction data, the task data of the hydraulic manipulator is corrected in a targeted manner. The main parameters to be corrected include the target position parameters, target posture parameters, and position parameters of the grasped object in the task. The adjustment of these parameters is intended to enable the hydraulic manipulator to perform tasks more accurately.

[0129] After the task data is corrected, the control parameters of the hydraulic manipulator are recalculated. New control parameters will be generated based on the corrected task data to ensure that the hydraulic manipulator can operate according to the corrected task requirements.

[0130] Finally, these modified control parameters are used to generate adaptive control commands for the hydraulic manipulator. These control commands will directly act on the control system of the hydraulic manipulator, guiding the manipulator to accurately complete the action according to the new task requirements.

[0131] Through the above steps, the method of the present invention can realize real-time monitoring, dynamic adjustment and optimization control of the hydraulic mechanical arm, thereby ensuring that the mechanical arm can maintain high precision and stability in different scenarios.

[0132] The technical solution proposed in this invention aims to solve the problem that when the hydraulic manipulator arm of the robot faces different working environments and tasks, the neural network adjustment process is delayed due to the handling of control problems with high uncertainty, thereby affecting the real-time operation accuracy of the hydraulic manipulator arm. The specific technical solution is as follows:

[0133] The present invention combines an adaptive neural network with the historical operating data of the hydraulic robotic arm to construct a control system that can adaptively adjust and predict the dynamic response of the robotic arm. It continuously optimizes and adjusts through multiple steps to adapt to the requirements of different working environments and tasks, thereby reducing the delay in the neural network adjustment process and improving the real-time operation accuracy of the hydraulic robotic arm.

[0134] Acquire the operating data (including status data, position, speed, acceleration, etc.) and task data (including target position, posture, position of grasped objects, trajectory planning parameters, etc.) of the hydraulic robot arm, and preprocess the data to prepare for subsequent neural network simulation and prediction.

[0135] Use an adaptive neural network to simulate the dynamic model of the robot arm and input the preprocessed data. The neural network predicts the dynamic response data of the robot arm in different states (such as position, velocity, and acceleration prediction data) based on the input data.

[0136] Obtain the motion prediction data of the robotic arm in different states for subsequent state evaluation and task data correction.

[0137] The historical operation data and task data are used to build the hydraulic manipulator operation status evaluation model. The real-time task data is substituted into the evaluation model, compared with the real-time operation data, and the operation error data is calculated.

[0138] The error data is split by task type and combined with historical data to form a data group to be optimized. The data group to be optimized is used to train and optimize the robot arm dynamics model and state evaluation model simulated by the adaptive neural network to reduce the prediction error.

[0139] Identify historical data to obtain the usage scenario type of the hydraulic manipulator. Process the real-time operation data according to the scenario type and use the optimized model for prediction. If the error between the predicted data and the real-time operation data is greater than the preset range, use the predicted data to correct the task data (such as adjusting the target position, posture and other parameters). Recalculate the control parameters based on the corrected task data and generate adaptive control commands.

[0140] The invention simulates the dynamic model of the mechanical arm by training an adaptive neural network to achieve accurate prediction of dynamic response. The state evaluation model and the optimization neural network are constructed using historical operation data and task data to improve the accuracy and real-time performance of the prediction. By calculating the operation error in real time and correcting the task data accordingly, the hydraulic mechanical arm can perform accurately in different working environments and tasks.

[0141] The technical solution of the present invention effectively solves the control delay problem of the hydraulic mechanical arm when facing different working environments and tasks by comprehensively applying adaptive neural networks, historical data analysis and real-time error feedback mechanisms, and improves the accuracy of real-time operations. The adaptive neural network control method for hydraulic mechanical arms provided by the present invention is of great significance for improving the intelligence level and application scope of hydraulic mechanical arms.

[0142] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations, and the above-described embodiments of the present invention do not constitute a limitation on the scope of protection of the present invention.

Claims

1. A hydraulic mechanical arm adaptive neural network control method, characterized in that: include: Step S101, obtaining hydraulic manipulator arm operation data and hydraulic manipulator arm task data, the hydraulic manipulator arm operation data includes hydraulic manipulator arm state data, hydraulic manipulator arm position, hydraulic manipulator arm speed and hydraulic manipulator arm acceleration, and the hydraulic manipulator arm task data includes target position, target posture, grasped object position, hydraulic manipulator arm trajectory planning parameters, hydraulic manipulator arm target torque, task constraint time and sensor data requirements; Step S102, preprocessing the hydraulic mechanical arm operation data and the hydraulic mechanical arm task data, using the preprocessed hydraulic mechanical arm operation data and the hydraulic mechanical arm task data to simulate the mechanical arm dynamics model, and predicting the dynamic response data of the mechanical arm in different states, to obtain the action prediction data of the mechanical arm in different states; Step S103, retrieve the historical operation data and historical task data of the hydraulic mechanical arm, build a hydraulic mechanical arm operation status evaluation model according to the historical operation data and historical task data of the hydraulic mechanical arm, obtain the real-time hydraulic mechanical arm operation data and real-time hydraulic mechanical arm task data, substitute the real-time hydraulic mechanical arm task data into the hydraulic mechanical arm operation status evaluation model, obtain the hydraulic mechanical arm operation status evaluation result, compare the hydraulic mechanical arm operation status evaluation result and the action prediction data of the mechanical arm in different states with the real-time hydraulic mechanical arm operation data, and obtain the real-time hydraulic mechanical arm operation error data; Step S104, splitting the real-time hydraulic mechanical arm operation error data according to the task type to obtain error data corresponding to each task, combining the error data corresponding to each task with the hydraulic mechanical arm historical operation data and the hydraulic mechanical arm historical task data in turn to obtain a data group to be optimized, and using the data combination in the data group to be optimized in turn to optimize the adaptive neural network simulation mechanical arm dynamics model and the hydraulic mechanical arm operation status evaluation model to obtain an optimized adaptive neural network simulation mechanical arm dynamics model combination and an optimized hydraulic mechanical arm operation status evaluation model combination; Step S105, identify the historical operation data and historical task data of the hydraulic mechanical arm, obtain the historical hydraulic mechanical arm usage scenario type, process the real-time hydraulic mechanical arm operation data according to the historical hydraulic mechanical arm usage scenario type, obtain the corrected hydraulic mechanical arm task data, and generate the hydraulic mechanical arm adaptive control command according to the corrected hydraulic mechanical arm task data.

2. The hydraulic mechanical arm adaptive neural network control method according to claim 1, characterized in that: The step S101 includes: A hydraulic mechanical arm data category is established based on historical hydraulic mechanical arm operation data and historical hydraulic mechanical arm task data. The hydraulic mechanical arm data category includes a hydraulic mechanical arm task category and hydraulic mechanical arm operation data corresponding to each hydraulic mechanical arm task category.

3. The hydraulic mechanical arm adaptive neural network control method according to claim 1, characterized in that: The step S102 includes: The adaptive neural network model is trained using historical hydraulic manipulator operation data and task data, and the pre-processed real-time hydraulic manipulator operation data and task data are input into the trained adaptive neural network simulation manipulator dynamics model. The adaptive neural network simulation manipulator dynamics model predicts the dynamic response data of the manipulator under different states based on the input data. The dynamic response data under different states include hydraulic manipulator position prediction data, hydraulic manipulator speed prediction data, and hydraulic manipulator acceleration prediction data. The motion prediction data of the robot arm in different states are obtained from the output layer of the adaptive neural network simulation robot arm dynamics model. The motion prediction data of the robot arm in different states are used for the operation state evaluation of the hydraulic robot arm and the correction of the task data.

4. The hydraulic mechanical arm adaptive neural network control method according to claim 2, characterized in that: The step S103 includes: The hydraulic manipulator arm historical operation data includes the hydraulic manipulator arm historical position, hydraulic manipulator arm historical speed, hydraulic manipulator arm historical force, hydraulic manipulator arm historical acceleration and hydraulic manipulator arm historical environmental parameters, and the hydraulic manipulator arm historical task data includes detailed information of each task of the hydraulic manipulator arm; The historical operation data and historical task data of the hydraulic mechanical arm are standardized according to the timestamp to obtain the historical operation data and historical task data of the hydraulic mechanical arm that are consistent in time; The average operation time characteristics, maximum load characteristics, and average failure interval characteristics are obtained from the historical operation data of the hydraulic manipulator; The historical operation data and historical task data of the hydraulic manipulator are used as training sets, and the recurrent neural network model is trained using the training set data to obtain the hydraulic manipulator operation status evaluation model.

5. The hydraulic mechanical arm adaptive neural network control method according to claim 4, characterized in that: The step S104 includes: By analyzing the identifiers and parameters in the task data, the task type corresponding to the real-time hydraulic manipulator operation data is obtained; The real-time hydraulic robot arm operation error data is classified according to the task type. For each task type, the task error data is extracted from the classified error data set, and the error data set corresponding to each task is output. The error data set corresponding to each task also includes a combination of the hydraulic robot arm historical operation data and task data.

6. The hydraulic mechanical arm adaptive neural network control method according to claim 5, characterized in that: The step S104 includes: For each task type, the real-time error data corresponding to each task type is combined with the historical operation data and historical task data of the hydraulic manipulator; Arrange the combined data according to a fixed data format and structure to form a data group to be optimized; Each data group to be optimized contains the hydraulic manipulator operation data, task data and corresponding error data under a specific task type; The adaptive neural network simulation mechanical arm dynamics model is trained using the data combination in the data group to be optimized to obtain the data combination training results in the data group to be optimized, and the parameters of the neural network are adjusted according to the data combination training results in the data group to be optimized to obtain the optimized adaptive neural network simulation mechanical arm dynamics model and the optimized hydraulic mechanical arm operation status evaluation model corresponding to each task; The optimized adaptive neural network simulation robot arm dynamics model and the hydraulic robot arm operation status evaluation model are combined to form an optimized model combination.

7. The hydraulic mechanical arm adaptive neural network control method according to claim 1, characterized in that: The step S105 includes: The historical operation data and historical task data of the hydraulic mechanical arm are identified to obtain the historical hydraulic mechanical arm usage scenario types, and a matching relationship is established between the hydraulic mechanical arm operation data corresponding to the historical hydraulic mechanical arm usage scenario types, the optimized adaptive neural network simulation mechanical arm dynamics model, and the optimized hydraulic mechanical arm operation status evaluation model to obtain the hydraulic mechanical arm usage scenario data group, and the real-time hydraulic mechanical arm operation data and the real-time hydraulic mechanical arm task data are matched with the hydraulic mechanical arm usage scenario data group to obtain the real-time hydraulic mechanical arm scene data group, and the real-time hydraulic mechanical arm scene data group is matched with the corresponding adaptive neural network simulation mechanical arm dynamics. The model and the hydraulic mechanical arm operation status evaluation model are used for data analysis to obtain the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario, and the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario are compared with the real-time hydraulic mechanical arm operation data. If the comparison result error is greater than the preset error range, the real-time hydraulic mechanical arm action prediction data for each scenario and the real-time hydraulic mechanical arm operation status prediction data for each scenario are used to correct the hydraulic mechanical arm task data to obtain the corrected hydraulic mechanical arm task data, and the hydraulic mechanical arm adaptive control command is generated according to the corrected hydraulic mechanical arm task data.

8. The hydraulic mechanical arm adaptive neural network control method according to claim 1, characterized in that: The step S105 includes: Compare the real-time hydraulic mechanical arm action prediction data of each scene, the real-time hydraulic mechanical arm operation state prediction data of each scene and the real-time hydraulic mechanical arm operation data, calculate the error value between the real-time hydraulic mechanical arm action prediction data of each scene and the real-time hydraulic mechanical arm operation data to obtain a first error value, calculate the error value between the real-time hydraulic mechanical arm operation state prediction data of each scene and the real-time hydraulic mechanical arm operation data to obtain a second error value; Determine whether the first error value and the second error value are greater than a preset error range. If the error value is greater than the preset range, use the real-time hydraulic mechanical arm action prediction data of each scene and the real-time hydraulic mechanical arm operation status prediction data of each scene as reference data, correct the target position parameters, target posture parameters and grasped object position parameters in the hydraulic mechanical arm task data to obtain the corrected hydraulic mechanical arm task data, recalculate the control parameters of the hydraulic mechanical arm according to the corrected hydraulic mechanical arm task data to obtain the corrected control parameters, and use the corrected control parameters to generate an adaptive control command for the hydraulic mechanical arm.